(September 3rd, 2014, 16:25)Tlaloc Wrote: What other discoveries did you make about what was hard or overpowered?
Too many to list. I collected data over the decades, but the data is still in raw format, which is not user friendly at the moment.
There are many people here who worked on balancing and I don't feel like doing that at this time. Balance and Speed and Insecticide by zitro1987 and Kyrub is my favorite.
In summary, for the expert player, MOM is a collection of useless features, units, and spells and ridiculously overpowered ones, which I can often exploit to the extreme. The original developers could not have possibly fixed those without years of expert player testing.
That's strange, it says your post I'm quoting below was posted Sept 3, 2014 14:20 but despite my regular checking, it didn't show up for me until an hour or so ago. I've subscribed to the thread now, so hopefully that will fix the problem.
(September 3rd, 2014, 16:20)WhiteMage Wrote:
(September 3rd, 2014, 15:01)Tlaloc Wrote: Could you please clarify #4?
List how many games had wizard with 0 of 1-pick skill (no 1-pick skill). How many games had exactly 1 of 1-pick skill. Exactly 2 of 1-pick skill. Etc. (E.g. Myrran is a 3-pick skill.) Clear? Also, exactly 1 of 1-pick skill with 1 of 2-pick skill…
Yes this is more clear, thanks. For this analysis would you want to include (for example) Runemaster as a 1-pick skill, Divine Power as a 2-pick skill, even though they cost more once the books are added in?
(September 3rd, 2014, 16:20)WhiteMage Wrote: If it is that fast, you shall go up to 1million. Power of 10 will allow easier human reading of results and spotting errors and outliers by eye. % will become easy to read, which is also a time saver.
Ok, 1million games it is. I agree power of 10 will be better.
(September 5th, 2014, 15:29)Tlaloc Wrote: would you want to include (for example) Runemaster as a 1-pick skill, Divine Power as a 2-pick skill, even though they cost more once the books are added in?
Good point. It is best if you do it both ways from the same generated games.
Ok, so if I break down your item #4 into the four sections below, will that suffice for your purposes? I'm using hypothetical numbers in the example below.
(Includes only 1-pick skills: Alchemy, Artificer, Charismatic, Conjurer, Mana Focusing)
Games with 0 1-pick skills: 44 (4%)
Games with 1 1-pick skills: 55 (5%)
Games with 2 1-pick skills: 44 (4%)
Games with 3 1-pick skills: 33 (3%)
Games with 4 1-pick skills: 22 (2%)
Games with 5 1-pick skills: 11 (1%)
(Includes all 1-pick skills)
Games with 0 1-pick skills: 44 (4%)
Games with 1 1-pick skills: 55 (5%)
Games with 2 1-pick skills: 44 (4%)
Games with 3 1-pick skills: 33 (3%)
Games with 4 1-pick skills: 22 (2%)
Games with 5 1-pick skills: 11 (1%)
Games with 6 1-pick skills: 6 (0.5%)
Games with 7 1-pick skills: 1 (0.1%)
(Includes only 2-pick skills: Channeler, Famous, Warlord)
Games with 0 2-pick skills: 44 (4%)
Games with 1 2-pick skills: 55 (5%)
Games with 2 2-pick skills: 44 (4%)
Games with 3 2-pick skills: 33 (3%)
(Includes all 2-pick skills)
Games with 0 2-pick skills: 44 (4%)
Games with 1 2-pick skills: 55 (5%)
Games with 2 2-pick skills: 44 (4%)
Games with 3 2-pick skills: 33 (3%)
Since Myrran is the only 3-pick skill, we can get the information you want from the list of individual ability stats, where it will say something like:
Games with Myrran: 144 (13%)
So I won't explicitly include 3-pick skills in the section above, unless you indicate otherwise.
Status update: I've been too busy this last month to devote much time to this project. Right now, I only have a few hours to work on it every couple of weeks, so I'd say it will take me another month or so to implement your algorithm.
You are on the right track. The key is to display the full list of 1-pick skills, 2-pick skills, etc. First, you ignore prerequisites, so report only 1-, 2-, 3-pick skills. For reading purposes, Myrran should be reported as a 3-pick skill, even if it is a set of 1. Also, you need to continue the statistics for "games with exactly 1 1-pick skill and 1 2-pick skill", etc. This is in the same row.
Once this is all done then consider prerequisites. So there will be 1-pick skills with 0 pick prerequisites, 1-pick skills with 1 prerequisite, 2 prerequisite, etc. Always give list of skills, which fall into that set. Some sets will be empty with no skill falling into.
All skills should fall into exactly 1 set for both calculations.
Ok, I've tried valiantly to figure out what you want and have come up with the attached sample stats file. Please review it and let me know if it's what you want. I didn't post it directly because it was getting quite long. I'm still not sure what you mean by "This is in the same row", or how many skill combinations (ie "Games with exactly 1 1-pick skill and 1 2-pick skill") you want listed, so if you have a preference there, let me know. If there are changes you like to make, I'd suggest modifying the file and attaching it so I know exactly what you'd like.
All the other stats, including normal standard deviations, are ready to go.
The text file sample is good. Regardless if it is impossible or it was just not generated by the program, do not print anything that has 0 occurrence. List everything that has >0 occurrence (out of 1million games). Keep this principle for everything. Not only for the combinations.
"This is in the same row",
You got this right in the text file. It was about printing them on the same row since they logically belong together.
"how many skill combinations (ie "Games with exactly 1 1-pick skill and 1 2-pick skill") you want listed, so if you have a preference there"
Again, list everything that has occurrence of >0. List nothing that has occurrence of 0 out of the 1million games. There should be no ambiguity about this. The sum of percentages should add up to 100 after successful generation (except rounding).
You also list counts and % for individual skills, etc.
I've pretty much finished implementing your algorithm. I quite like the games it is generating. I encountered a case where I had to tweak your algorithm a bit. The case is rather remarkable:
We are in Book Favor mode and have to pick a skill. There is now no skill that can be picked to use up the 1 remaining pick. I tweaked the algorithm to pick a book if it ever gets stuck like this. I still have to integrate your algorithm into my stats program some more before I can give you the stats on it. I hope to get that out in a week or so when I have some more time to work on it.
Of course. Good choice. In my pseudocode I stated
"Once skill is selected, go to Spend Picks." This can be interpreted/extended to apply to the case when it is impossible to do it. (Here and everywhere else in general). Deadlock must be resolved. Simply consider the condition met and move to next line. I agree that it is more efficient to not go to Spend Picks in your stated case, since it may waste even more time here, but would not deadlock though.
The pseudocode was written with the manual fast execution mindset. A computer program has to cover all cases. This is why it is good that you generate a million samples so things like this come out empirically.
This was kind of similar to trying to take death book, once you already have life book, which is impossible, but my code does not care. I though about it when I wrote it, but again simple code and fast manual execution was my goal. Checking for all prerequisites would have taken too much time to write code for and execute with very little benefit. Simplicity is a good design and programming principle.
Keep going! Thanks,
Ok, thanks for the encouragement. I still enjoy working on this project so the motivation has been there. I agree it's good to make things as simple as possible.
At long last, here are the stats for your algorithm. The average number of book is 7.95 which is good. As expected, the number of life and death books generated is low compared to the other book types, and the occurrence of skills generally decreases as they become more expensive, with Divine Power and Infernal Power being the least frequent.
Also, the "Average number of occurring books" stat measures the average number of books in games where there was at least 1 of that book type.
I hope you find the stats useful. When time allows, I'm planning to generate stats for your algorithm when it uses the PickWinner heuristic for picking books instead of the Reroll heuristic, to see how the two compare. The PickWinner heuristic would pick the winning book type (either life or death) at the start and then put all life and death book picks into the winning book type instead of rerolling.
Let me know if you'd like to make any changes to your algorithm based on these stats.
